Full Terms & Conditions of access and use can be found athttps://www.tandfonline.com/action/journalInformation?journalCode=ueqe20

Journal of Earthquake Engineering

ISSN: 1363-2469 (Print) 1559-808X (Online) Journal homepage: https://www.tandfonline.com/loi/ueqe20

Implementation of Bond-Slip PerformanceModels in the Analyses of Non-Ductile ReinforcedConcrete Frames Under Dynamic Loads

Jiuk Shin, Lauren K. Stewart, Chuang-Sheng Yang & David W. Scott

To cite this article: Jiuk Shin, Lauren K. Stewart, Chuang-Sheng Yang & David W. Scott(2017): Implementation of Bond-Slip Performance Models in the Analyses of Non-DuctileReinforced Concrete Frames Under Dynamic Loads, Journal of Earthquake Engineering, DOI:10.1080/13632469.2017.1401565

To link to this article: https://doi.org/10.1080/13632469.2017.1401565

Published online: 27 Dec 2017.

Submit your article to this journal

Article views: 131

View Crossmark data

Implementation of Bond-Slip Performance Models in theAnalyses of Non-Ductile Reinforced Concrete Frames UnderDynamic LoadsJiuk Shin, Lauren K. Stewart, Chuang-Sheng Yang, and David W. Scott

School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA USA

ABSTRACTNon-ductile reinforced concrete frames have seismic vulnerabilitiesdue to inadequate reinforcing details. Adequately characterizing theperformance of these details is critical to the development of analy-tical models. This study presents a methodology to simulate theresponse of such frames with and without fiber-reinforced polymercolumn jackets. The bond-slip effects between reinforcing bars andsurrounding concrete, observed in column lap-splice and beam–col-umn joints, are modeled with one-dimensional slide line models inLS-DYNA. The model is defined from failure modes and bondingconditions observed in full-scale dynamic tests and can predictstory displacements, inter-story drifts, and damage mechanisms.

ARTICLE HISTORYReceived 24 November 2016Revised 22 August 2017Accepted 2 November 2017

KEYWORDSNon-Ductile ReinforcedConcrete Frame;Fiber-Reinforced PolymerColumn Jacket; FiniteElement; Bond-Slip Effect;Full-Scale Dynamic Test

1. Introduction

Reinforced concrete (RC) building structures constructed before the 1970s were typicallydesigned only for gravity loads. This design practice led to seismically deficient detailing incolumns and beam–column joints, such as small-diameter and largely spaced transversereinforcement in columns, insufficient length of column lap-splice bars, and inadequate detail-ing of beam–column joints (e.g. no transverse reinforcing bars in panel zones, straight ancho-rage of positive beam reinforcing bars, and discontinuous positive beam reinforcing bars)[Bracci et al., 1995; El-Attar et al., 1997; Priestley, 1997; Wright, 2015]. Such inadequatereinforcement details can result in a poor bond-slip condition between the reinforcing barsand the surrounding concrete observed in both column lap-slices and beam–column joints.Post-earthquake reconnaissance reports [Aschheim et al., 2000; Sezen et al., 2000] have demon-strated that inadequate reinforcement details in columns and beam–column joints can lead topremature collapse of the structure. For example, the 1999 Kocaeli earthquake in Turkey led tothe collapse ofmultiple RC building structures designedwithout appropriate seismic detailing ofreinforcement. Many of these collapses were attributed to shear and/or lap-splice failures incolumns, in addition to the development of significant damage in beam–column joints.

To prevent the premature failure of non-ductile RC frame structures, seismic retrofittechniques for RC columns using fiber-reinforced polymer (FRP) composites have beenwidely used [Seible et al., 1997; Xiao et al., 1999; Haroun et al., 2003; Sause et al., 2004;

CONTACT Lauren K. Stewart lauren.stewart@ce.gatech.edu School of Civil and Environmental Engineering,Georgia Institute of Technology, Atlanta, GA 30332, USA.Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/ueqe.

JOURNAL OF EARTHQUAKE ENGINEERINGhttps://doi.org/10.1080/13632469.2017.1401565

© 2017 Taylor & Francis Group, LLC

ElGawady et al., 2010; Yan and Pantelides, 2011; Shin et al., 2016]. These previous experi-mental studies have demonstrated that FRP column jackets can enhance the flexural, shear,and lap-splice capacities of non-ductile RC columns as well as increase the bucklingresistance of reinforcing bars by providing additional confining pressure. A recent study[Shin et al., 2016] performed a series of seismic dynamic tests for a full-scale, two-story non-ductile RC test frame, designed for only gravity loads, in non-retrofitted (referred to as “as-built test frame”) and retrofitted (referred to as “retrofitted test frame”) configurations, toinvestigate the dynamic responses under seismic loading. For the retrofitted test frame,prefabricated FRP column jackets were installed on the first-story columns in the full-scaletest frame as a retrofit system. By comparing the dynamic responses between the as-builtand retrofitted test frames, the effectiveness of the retrofit system was demonstrated in termsof inter-story drifts, column rotations, and drift concentration factors (DCFs).

This paper aims to validate the dynamic responses of full-scale, non-ductile, andretrofitted test frames in terms of damage, story displacements, and inter-story driftsusing finite element (FE) frame models. The FE frame models are developed with thenonlinear explicit analysis program, LS-DYNA [2013]. The numerical models are verifiedwith the measured responses from a series of full-scale dynamic tests [Wright, 2015; Shinet al., 2016]. Based on the dynamic responses measured from the full-scale tests, this studyidentifies true bond-slip behavior in individual column lap-splice and panel zones for boththe as-built and retrofitted test frames. To represent the bond-slip behavior observed inthe full-scale dynamic testing, a modeling methodology that characterizes bond-slipperformance using LS-DYNA is proposed and incorporated into the FE frame models.This empirical bond-slip performance behavior is simulated with a one-dimensional slideline model. The modeling methodology for representing the bond-slip performance isapplicable for implementation in analyses of other structures with similar, seismicallydeficient detailing. Additionally, the effect of changing parameters on the bond-slipperformance of the FE frame models is investigated. Previous works involving FE analysison RC frames usually focused on the validation of experimental responses under a quasi-static monotonic or cyclic load at the element or basic system level, such as non-ductileand FRP-confined RC columns [Kwon and Spacone, 2002; Hu et al., 2003], and RC beam–column joints [Deaton, 2013]. This study verifies the proposed modeling methodology atthe frame level with the full-scale dynamic experimental results.

2. Full-Scale Dynamic Testing of As-Built and Retrofitted RC Test Frames

2.1. Description of Full-Scale Test Frames and Test Setup

A series of full-scale dynamic tests were performed on four identical two-story, two-bay, non-ductile RC test frames constructed at the Georgia Institute of Technology (see Figure 1a). Onetest frame was constructed and tested in an as-built configuration, and one test frame wasretrofitted with prefabricated FRP column jackets on the first-story columns (see Figure 1b).The identical RC test frames were designed in accordance with ACI 318-63 [1963]. Since thisbuilding code has no seismic requirements, the test frames were composed of inadequatereinforcement details in columns and beam–column joints. The design included short columnlap-splice lengths, poor confinement, and weak column strong beam (WCSB) systems. Theseismically deficient reinforcement details of the RC test frames are summarized in Figure 2.

2 J. SHIN ET AL.

To enhance the ductility, shear, and lap-splice capacities of the as-built test frame, an FRPjacket system was designed using the retrofit design process proposed by Seible et al. [1997].Additionally, as specified in Federal Emergency Management Agency (FEMA) 547 [2006],gaps of approximately 13 mm were left at the top and bottom of the column to prevent theinteraction between the FRP jacket and the adjacent elements (e.g. slab, beam, and founda-tion). The test frames were excited beyond their linear elastic limit using a mobile shaker

(a) (b)

(c)

Figure 1. Full-scale, two-story two-bay non-ductile RC test frames: (a) four identical full-scale testframes; (b) FRP jacketed column in the retrofitted test frame; (c) hydraulic linear shaker on the roof ofthe RC test frames [Shin et al., 2016].

Figure 2. Reinforcement details of the full-scale, as-built test frame: (a) first-story column; and (b) first-story exterior beam–column joint [Wright, 2015].

JOURNAL OF EARTHQUAKE ENGINEERING 3

system [NEES@UCLA, 2015] installed on the roof of the test frames as shown in Figure 1c.The mobile shaker sequentially applied seismic (1940 El Centro earthquake) and sine motionsto the test frames. Story displacements, hinge rotations, accelerations, and bar strains inducedby the top excitation weremeasured. Additional information related to the experimental setupof the as-built and retrofitted test frames can be found inWright [2015] and Shin et al. [2016],respectively.

2.2. Experimental Results

At the completion of the dynamic loading sequences, the as-built test frame was severelydamaged in the first-story columns and exhibited mechanisms such as vertical splitting cracksalong the column lap-splice bars and shear cracks within the column lap-splice regions. Nodamage was detected in the second story. The lap-splice failure in the first-story column basesand the pullout failure in the first-story exterior beam–column joints were demonstrated by themeasured relation between the hinge rotations and the reinforcing bar strains [Wright, 2015].Such bond-slipmechanisms significantly contributed to the soft-story behavior of the first story.

3. Methodology for the FE Model

3.1. Structural Geometry Modeling

A methodology for modeling the RC frame structure with bond-slip effects was developedand verified with the experimental responses measured by the full-scale dynamic experi-ments. The test frames were modeled using the nonlinear, explicit FE software LS-DYNA[LSTC, 2013]. Figure 3 shows a three-dimensional view of the as-built and retrofitted FEframe models. As shown in Figure 3a, the FE frame models were developed using a half-symmetry condition, which was used to reduce computational demands. To enforce thecondition on the plane of symmetry, the two rotational degrees-of-freedom parallel to thesymmetry plane (Rx and Rz in the global coordinate system) and the translational degree-of-freedom perpendicular to the symmetry plane (Dy in the global coordinate system)were restrained. Additionally, the foundation bases were restrained in all translational androtational directions, which simulated the fixed condition. Live loads in the experimentwere simulated by placing steel rails on the second and third floors of the frame. Theseweights were converted to masses, and then equally distributed as nodal masses on eachelement of the floor slab using the option Element_Mass in LS-DYNA.

The concrete column and beam models in the longitudinal direction (x-direction)utilized eight-node solid elements with single point integration. All reinforcing barswere modeled using two-node Hughes–Liu beam elements. The longitudinal and trans-verse reinforcing bars in the column and beam elements were connected to the concretemesh nodes. The nodes that linked the concrete and reinforcement mesh were shared.These shared nodes were fully bonded; however, to capture bond-slip effects, the lap-splicebars in the columns and straight anchorages in the joint regions have separate nodes fromthe concrete mesh nodes. These separate nodes were linked with one-dimensional slidelines. Since the foundation, slab, and transverse beam had no significant damage duringthe dynamic tests, those elements were modeled using shell elements with an elasticmaterial model (MAT001 in LS-DYNA) to develop a more efficient model. The elastic

4 J. SHIN ET AL.

material model was reduced by stiffness reduction factors specified in ASCE 41-13 [2014].To impose the shaker forces measured from the full-scale dynamic testing, a shaker platewas placed on the top of the FE frame model. The shaker plate was represented using solidelements with a rigid material (MAT020 in LS-DYNA).

Figure 3b shows the model of the FRP column jacketing system on one of the first-story columns, consisting of non-shrink grout and FRP composite. The non-shrinkgrout model utilized solid elements to provide additional confining pressures. TheFRP jacket model, placed on the surface of the grouting model, was modeled usingthe shell elements with the corresponding thicknesses to the FRP jackets of the retro-fitted test frame: two-layer FRP jacket (1.32 mm in thickness) on the top and bottom ofthe column, and one-layer FRP jacket (0.66 mm in thickness) in the middle of thecolumn [Shin et al., 2016]. To represent the half-symmetry condition for the retrofittedFE frame model, the FRP jacket and grout elements were restrained in the y-transla-tional direction and the x- and z-rotational directions. Additionally, as illustrated inFigure 3b, the FRP jacket system was assumed to have two different interface layers: acontact layer between the concrete and non-shrink grout models and a contact layerbetween the non-shrink grout and FRP jacket models. These interface layers weresimulated using the LS-DYNA function CONTACT AUTOMATIC SURFACE-TO-SURFACE (referred to as “surface-to-surface contact”), which was developed by Tabieland Wu [Tabiei and Wu, 2000]. In this contact function, the frictional coefficients of theinterface layers were assumed to be 0.8 based on previous works [Moradi, 2007;Davidson, 2008; Lee and Shin, 2016] that recommended the use of coefficient of frictionvalues ranging from 0.6 to 0.9 under dynamic loading. The coefficient of friction valueof 0.8 was incorporated into both interface layers between the concrete and non-shrinkgrout models and between the grout and FRP jacket models.

Figure 3. Three-dimensional view of the FE frame models: (a) as-built FE frame model; (b) first-storycolumn of the retrofitted FE frame model.

JOURNAL OF EARTHQUAKE ENGINEERING 5

3.2. Material Models

3.2.1. Concrete MaterialTo predict the concrete behavior, LS-DYNA provides several material models, such asWINFRITH_CONCRETE (MAT084, often referred to as the “Winfrith model”)[Broadhouse, 1986], CSCM (MAT159, referred to as the “CSC model”) [Schwer and Murray,1994], and CONCRETE_DAMAGE_REL3 (MAT072R3) [Malvar et al., 1997; Crawford et al.,2012], which is well known as the Karagozian & Case (K&C) concrete model (referred to as the“KCCmodel”). Thesemodels have a default parameter generation function (e.g. the unconfinedcompressive strength of the concrete) and capture post-peak strain softening, shear dilation, andconfinement effect in concrete behavior. Wu et al. [2010] examined the triaxial behavior of asingle solid element modeled with three different concrete constitutive models, and comparedthe simulated results with triaxial compression tests. The numerical study demonstrated that theKCC model can reproduce concrete damage behavior, softening, modulus reduction, sheardilation, and confinement effect under the wide range of confining pressures. Additionally, thisKCCmodel has been extensively verified with the experimental responses obtained from quasi-static, blast, and high-velocity impact loading tests [Wu and Crawford, 2015]. Therefore, thisstudy selected the KCC model to simulate the concrete behavior among the various concreteconstitutive models provided in LS-DYNA.

The KCC model is characterized by three independent shear failure surfaces: themaximum surface (Δσm), the yield surface (Δσy), and the residual surface (Δσr). The strainhardening-softening responses in axial stress–strain behavior are established by the com-bination of three independent shear surfaces and damage function as given in Equations(1) and (2), where σ is axial stress, P is hydrostatic pressure, η(λ) is damage function,λ is the effective plastic strain, and λm is the effective plastic strain corresponding to Δσm[Malvar et al., 1997; Wu and Crawford, 2015].

σ ¼ ηðλÞ � ½ΔσmðPÞ � ΔσyðPÞ� þ ΔσyðPÞ strain hardeningð Þ λ � λm (1)

σ ¼ ηðλÞ � ½ΔσmðPÞ � ΔσrðPÞ� þ ΔσrðPÞ strain softeningð Þ λ> λm (2)

The KCCmodel can simulate a shear dilation using a parameter, ω. This parameter can capturethe expansion of concrete as it cracks. If high confinement effects are present in FRP jackets, theω parameter can contribute to providing a confining pressure, and thus it can increase thestrength and ductility [Crawford et al., 2012]. In other words, the ω parameter plays a criticalrole in simulating reasonable confinement effects. Crawford et al. [2013] suggested ω = 0.9 forwell-confined concrete components (FRP-jacketed RC column) and ω = 0.5 or 0.75 for poorlyconfined concrete components (non-ductile RC column). Based on these past studies, this studyemployed the ω parameter of 0.9 for the retrofitted columns and the ω parameter of 0.5 for theas-built columns. Table 1 shows themainmaterial parameters of the KCCmodel for the as-builtand retrofitted FE framemodels. The tensile strength of the concretematerials was computed bythe default parameter generation function of the KCC model.

3.2.2. Steel MaterialPLASTIC_KINEMATIC (MAT003, referred to as “elasto-plastic material model”) hasbeen widely used to simulate steel behavior. The steel stress–strain curve of the elastic-plastic material model represents bilinear behavior with linear isotropic hardening. The

6 J. SHIN ET AL.

parameters of this material model were defined based on the quasi-static testing results foreach rebar size embedded in the test frames. The yield strengths and elastic modulus forthe ϕ10, ϕ19, and ϕ25 steel reinforcing bars are summarized in Table 2. The straincorresponding to the ultimate strength was assumed to be approximately 15%. After theultimate strength, the steel material models represented softening behavior until 25%strains. As shown in Figure 4, the selected material model was verified with the materialtests for the ϕ10, ϕ19, and ϕ25 steel reinforcing bars.

3.2.3. FRP Composite MaterialThis study modeled the prefabricated FRP jackets on the first-story columns using theORTHOTROPIC_ELASTIC (MAT002, referred to as “orthotropic material”) model in LS-DYNA. This material model is characterized by elastic modulus (E), shear modulus (G), andPoisson’s ratio (ν) in terms of the three local principal axes (a, b, and c). Thus, the orthotropicmaterial can simulate the directional characteristics of FRP materials [LSTC – LivermoreSoftware Technology Corporation, 2013]. The main parameters (Ea, Gab, and νab) of theorthotropic material in the hoop direction used in this study are 95.5 GPa, 4.5 GPa, and 0.28,respectively. The failure criterion was assigned 1.1% ultimate strain of the FRP material.Recent studies [Mutalib and Hao, 2010; Nam et al., 2009; Youssf et al., 2014] utilized the

Table 1. Main material parameters of the KCC model.As-Built FE frame Retrofitted FE frame

Story levels ElementConcrete

strength (MPa) ω-parameterConcrete

strength (MPa) ω-parameter

First story Column 31.5 0.5 32.8 0.9Beam 25.0 0.5 26.5 0.5

Second story Column 28.5 0.5 30.3 0.5Beam 23.5 0.5 23.5 0.5

Table 2. Main parameters of the elasto-plastic material model.

RebarYield strength

(MPa)Ultimate strength

(MPa)Elastic modulus

(GPa)

ϕ10 (Diameter = 10 mm) 506.4 738.6 196.5ϕ19 (Diameter = 19 mm) 431.7 734.4 193.8ϕ25 (Diameter = 25 mm) 512.5 663.3 208.5

Figure 4. Comparison of steel stress–strain behavior between experiment and simulation: (a) ϕ10 steelrebar; (b) ϕ19 steel rebar; and (c) ϕ25 steel rebar.

JOURNAL OF EARTHQUAKE ENGINEERING 7

orthotropic model to simulate and verify the FRPmaterial behavior for FRP-strengthened RCstructures. In particular, Youssf and ElGawady [2014] verified the material behavior of FRP-confined concrete by using the orthotropic model for the FRP materials, which indicates thatthe orthotropic material can predict the confining pressure provided by the FRP materials.

3.3. Bond-Slip Modeling

The bond-slip effects between steel reinforcing bars and concrete significantly affected thestructural responses of RC structures [Spacone and Limkatanyu, 2000; Luccioni et al., 2005;Bao and Li, 2010]. The non-ductile RC frame has short lap-splice reinforcing bars in thecolumns, a straight anchorage of positive (bottom) beam reinforcing bars, and no transversereinforcement in the beam–column joint regions. Since such inadequate reinforcing detailscan lead to poor bonding conditions in RC structures, the bond-slip effects are critical todevelop the as-built FE frame models. Several existing RC building structures were built withinadequate concrete cover using lower-quality concrete materials that were available at thetime of construction; these factors often lead to corrosion of steel reinforcement.Reinforcement corrosion can result in a significant loss of bond strength between thesurrounding concrete and the bar, decreasing the overall seismic performance of the struc-ture [Deaton, 2013]. For RC structures, the effect of corrosion on the bond-slip behavior istypically critical after the onset of spalling in the concrete cover [Beeby, 1978]. The experi-mental results used to verify the FE frame model in the present work came from tests on non-damaged test frames performed within 18 months of construction. As such, steel corrosionwould have no significant impact on bond-slip behavior. Therefore, the FE frame modelspresented here do not include the effect of steel corrosion on bond-slip behavior.

3.3.1. One-Dimensional Slide Line ModelA one-dimensional slide line model provided by LS-DYNA can transfer interfacial shearforces between the slave nodes of the reinforcing bar beam elements and the master nodes ofthe concrete solid elements. The interfacial forces are proportional to the slip displacementbetween the slave nodes and the master nodes. These one-dimensional slide line models cansimulate the bond-slip effects by defining the bond shear modulus (Gs), maximum elastic slips(smax), and the damage curve exponential coefficient (hdmg), as given in Equation (3). The hdmg

decays the bond shear stress with the increment of plastic slip displacement (Δsp), andD is thedamage parameter, the sum of the absolute values of Δsp (Dn+1 = Dn + Δsp, where n is anincremental step). As defined in Equation (3), the bond-slip behavior is assumed to be bilinearand the bond stress deterioration is initiated after reaching τmax (= Gs∙s).

τ ¼ Gss; s � smax

τmaxe�hdmgD; s> smax(3)

Shi et al. [2008] modeled the bond-slip interface behavior between the beam elements andthe surrounding concrete solid elements in an RC column using the one-dimensional slideline in LS-DYNA for blast loading. The FE column model with bond-slip effects showsbetter prediction of the blast responses than the numerical model calibrated by Woodsonand Baylot [1999]. Researchers [Shi et al., 2009] also modeled the bond-slip effects withthe one-dimensional slide line model to simulate pullout responses in RC beam elementsand verify the responses with experimental results. The bond-slip parameters in the one-

8 J. SHIN ET AL.

dimensional slide line model were calibrated based on the experimental results. Thecalibrated FE models demonstrated the pullout testing results. Additionally, to demon-strate the collapse scenarios of RC beam–column assemblies, Bao et al. [2008] developedFE models with the bond-slip effects. These bond-slip effects were also simulated withone-dimensional slide line models between beam longitudinal bars and the surroundingconcrete in the panel zone. The FE models appropriately predict the collapse scenarios.

3.3.2. Experimental Response for Bond-SlipA previous experimental study on a non-ductile RC test frame [Wright, 2015] demonstratedbond-slip behavior in column lap-splice and straight anchorage (positive beam reinforcingbars in beam–column joint areas) using test results obtained from a full-scale dynamicexperiment. The bond-slip behavior in lap-splice and joint areas was identified by compar-ing the measured hinge rotations in beams or columns to reinforcing bar strain in thoseareas under dynamic loads. The relationship between the peak hinge rotations and thecorresponding bar strains is summarized in Figures 5 and 6, respectively, for each loadingsequence for the as-built and retrofitted test frames. Figure 5a shows the relationshipbetween peak column hinge rotations in first-story column bases and the correspondingbar strains for each loading sequence. If no bond-slip failure in the lap-splice regions wasobserved, then the peak column hinge rotations should continuously increase in accordancewith the shaker force increment. Furthermore, the bar strains should also keep increasingprior to the yielding of steel reinforcing bars. However, it was observed in the experimentthat the lap-splice bar strain decreased after reaching the peak bar strain, which occurredbefore the yielding of steel reinforcing bars. This phenomenon is believed to have occurreddue to concrete cracks in the lap-splice regions (e.g. splitting cracks along the columnreinforcing bars and shear cracks), which contributed to the losses of the interface forcesbetween the lap-splice bars and the surrounding concrete. Figure 5b demonstrates thepullout behavior in the positive beam reinforcing bars in the first-story exterior beam–column joints. However, the maximum bar strains in the negative beam reinforcement arecontinuously increased due to 180º anchorage hooks. More detailed test results of the bond-slip behavior for the as-built test frame can be found in Wright [2015].

Figure 5. Relationships between peak hinge rotations and bar strains for the as-built test frame: (a)first-story column; and (b) first-story exterior beam–column joint.

JOURNAL OF EARTHQUAKE ENGINEERING 9

Through the full-scale dynamic testing of the retrofitted test frame, the bond-slipresponses were also observed using the same approach as the as-built test frame.Figure 6a proves that the installation of the FRP column jacket system in first-story columnscan help delay the bond-slips of the column lap-splice bars since the maximum bar strainskeep increasing with the peak column hinge rotations. This is due to the additionalconfining pressures provided by the FRP jacket system. As shown in Figure 6b, strainreadings in bottom reinforcing bars taken near an exterior beam–column joint indicated adecrease in bar strains after they reached a maximum value but before yielding in the bar.This decrease indicates a pullout failure in the exterior beam–column joints. Based on theseobservations in the full-scale dynamic tests, this study determined the bonding conditionsand failure modes for the locations where the bond-slip effects were significantly affected.

3.3.3. Numerical Bond-Slip ModelThis study characterized the bond-slip performance in the FE frame models using theCEB-FIP MODEL CODE [1990]. The model code defines the bond stress–slip relationshipdepending on failure modes (e.g. splitting (lap-splice) and pullout failure modes) andbonding conditions (e.g. good and poor conditions). The bonding conditions can bedetermined by confining pressures regarding the concrete cover and the reinforcementdetailing, such as column lap-splice length, transverse reinforcing details, and anchoragedetails in beam–column joints. Figure 7 compares the bond stress–slip relationshipsbetween the model code and the one-dimensional slide line model in LS-DYNA. Thebond stress deterioration was captured using the hdmg parameter discussed in Section3.3.1. The values of hdmg in terms of the failure modes and bonding conditions were foundthrough calibrating the residual bond stresses of the model code with those of the one-dimensional slide line model. The values of hdmg for good and poor bonding conditions inthe lap-splice failure are 0.25 and 0.065, respectively, and the value of hdmg for good andpoor bonding conditions in the pullout failure is 0.01.

Figure 8 shows the representative bond-slip modeling used in the FE frame models withrespect to the failure modes and bonding conditions determined based on the experi-mental studies [Wright, 2015; Shin et al., 2016]. Figure 9 indicates the locations wherebond-slip effects are thought to occur. As shown in Figure 8a and b, the bond-slip models

Figure 6. Relationships between peak hinge rotations and bar strains for the retrofitted test frame: (a)first-story column; and (b) first-story exterior beam–column joint.

10 J. SHIN ET AL.

Figure 7. Comparisons of bond stress–slip relations between the CEB-FIP Model Code and the one-dimensional slide line model in LS-DYNA: (a) splitting failure and poor bond; (b) splitting failure andgood bond; (c) pullout failure and good bond; and (d) pullout failure and poor bond.

Figure 8. Examples of bond-slip modeling for failure modes and bonding conditions in LS-DYNA: (a)first-story as-built column; (b) first-story exterior beam–column joint; (c) first-story retrofitted column;(d) second-story column; and (e) second-story interior beam–column joint.

JOURNAL OF EARTHQUAKE ENGINEERING 11

in first-story as-built columns and positive beam reinforcing bars were defined based onthe full-scale dynamic testing results (see Figures 5 and 6). The test results also demon-strated that the installation of FRP column jackets system improved the bonding conditionof the column lap-splice bars (see Figure 6a). Thus, as shown in Figure 8c, the bond-slipeffects in the column lap-splice regions of the retrofitted FE frame model were simulatedwith the bond stress–slip response determined by the splitting failure mode and the goodbonding condition. The second-story columns have longer lap-splice length and smallerspacing of column ties with seismic detailing (i.e. 135º specified angle), which was totransfer the vibration loads from the second story to the first story without an unexpectedfailure during the experiment [Shin et al., 2016]. Thus, the bond-slip behavior of lap-splicezones in second-story columns was modeled with a splitting failure mode and goodbonding condition, as given in Figure 8d.

Since no instrumentation was installed inside the panel zones and transverse beamssupporting slabs inhibited a visible inspection on the surface of panel zones, this studyassumed failure modes and bonding conditions between the column reinforcing bars and thesurrounding concretes inside the panel zones. Past experimental studies [Akguzel, 2011;Engindeniz, 2008; Park andMosalam, 2012] on non-ductile beam–column joints constructedprior to the 1970s, which have no transverse reinforcing bars inside the panel zones, detectedvisible damage on the surface of panel zones, such as shear cracks and splitting cracks alongthe column reinforcing bars in the panel zones. In particular, the splitting cracks resulted inconfinement losses between the column reinforcing bars and the surrounding concrete, andthese confinement losses can produce significant bond-slip effects. The second-story panelzones, as given in Figures 8e and 9, have transverse beam reinforcing bars, similar to currentseismic design requirements. These can minimize the bond-slip effects of the columnreinforcing bars inside the panel zones. Therefore, this study assumed bond-slip conditionsfor the column reinforcing bars in the panel zones as the splitting failure mode with poorbonding condition in the first story (see Figure 8b) and the splitting failure mode with good

Figure 9. Bond-slip model locations for the as-built FE frame model.

12 J. SHIN ET AL.

bonding condition in the second story (see Figure 8e). This study utilized the FE framemodels with all possible bond-slip effects for the verification with the experimental responses.

4. Verification of FE Numerical Frame Models

4.1. Shaker Forces

During full-scale dynamic testing, test frames were vibrated by shaker forces induced from ahydraulic linear shaker on the roof. The linear shaker generated twodifferent types of excitations:seismic and sine vibrations. The amplitudes of the shaker force in each input excitation werescaled by the increase in the target displacement of the linear shaker [Shin et al., 2016]. To verifythe elastic and inelastic responses, this study selected two different types of loading scenarios foreach FE frame model: the 1940 El Centro (EC) earthquake with 203 mm (8 inch) targetdisplacement (EC 8) and double sine pulse (DSP) vibration with 660 mm (26 inch) targetdisplacement (DSP26–ultimate loading scenario) for the as-built FE framemodel, andEC8 andDSP vibration with 508 mm (20 inch) target displacement (DSP 20 – ultimate loading scenario)for the retrofitted FE frame model. The full-scale dynamic testing for the as-built test frame wasperformed under various loading sequences. Among them, DSP 26 was selected as the ultimateloading scenario for the as-built FE frame model because it induced significant bond-slip effectsin the column lap-splice and panel zones during the full-scale dynamic testing for the as-built testframe [Wright, 2015]. For the retrofitted FE frame model, DSP 20 was chosen as the ultimateloading scenario, which initiated pullout failure between the bottom beam reinforcing bars andthe surrounding concrete. The seismic and sine vibrations were applied to the rigid plates of theFE framemodels as shaker forces. These shaker forces, FðtÞ, were computed as given in Equation(4), wherems is the mass of the linear shaker and €xsðtÞ is the absolute acceleration of the shaker.This acceleration datawasmeasured by an accelerometer,mounted directly to the shakermass inthe in-plane direction. To eliminate the noises of themeasured acceleration, the acceleration datawere filtered. As an example, the filtered and measured shaker accelerations for the as-built testframe are plotted in Figure 10, with EC 8 in Figure 10a and DSP 26 in Figure 10b.

FðtÞ ¼ �ms€xsðtÞ (4)

4.2. As-Built FE Frame Model

Figure 11 compares the displacement-time history responses between the experimental andsimulated results in the first and second stories under the EC 8 loading. As observed in the full-scale dynamic tests, this loading scenario did not induce any significant damage on the structure[Wright, 2015]. Thus, the bond-slip effects were expected to bemarginal in this loading scenario.This loading scenario was taken into account to verify the dynamic behavior within an elasticrange. The as-built FE frame model predicted the experimental results in terms of the responseperiod over the full range of time. The maximum absolute story displacement of the experi-mental and simulated responses at t1 (= 1.33 s) and t2 (= 4.19 s) is plotted in Figure 12a and b,where t1 and t2 denote the time at which the first and second stories have the maximumdisplacements in the displacement-time history responses obtained from the experiment,respectively. The FE frame model slightly underestimates the maximum displacement at t1 byapproximately 10.0% (see Figure 12a), and it overestimates the maximum displacement at t2 by

JOURNAL OF EARTHQUAKE ENGINEERING 13

approximately 9.0% (see Figure 12b). Figure 13 shows the peak inter-story drift ratios obtainedfrom the experiment and simulation. The peak inter-story drift ratios of the simulated results areplotted at the time when themaximum drift response occurs for each story level.While the peakinter-story drift ratio in the first story is approximately 4.0% lower than the experimental results,the peak inter-story drift ratio in the second story is approximately 10.0% higher than theexperimental results.

Figure 10. Measured and filtered shaker accelerations for the as-built FE frame model: (a) seismicexcitation, 1940 El Centro earthquake (EC 8); and (b) double sine pulses (DSP 26).

Figure 11. Comparison of time-history responses for the as-built FE frame model between experimentaland simulated responses: (a) first story; (b) second story.

14 J. SHIN ET AL.

Figure 14 illustrates the experimental and simulated time history responses in the firstand second stories under the DSP 26 loading. Overall, the simulated and experimentalresults are in good agreement in terms of the response periods. The maximum absolutestory displacements for the first and second stories were found at t1 = 11.4 and t2 = 11.5 s,respectively. The story displacements at t1 and t2 are plotted in Figure 15a and b,respectively. The maximum simulation variation for the story displacements of the FEframe model at t1 and t2 is approximately 7.0% and 8.0%, respectively. Figure 16 comparesthe peak inter-story drift ratios between the experiment and the simulation. The peakinter-story drift ratio in the second story was overestimated by approximately 9.0%.

Overall, the simulation variation was estimated below 10.0%. This variation was attributed tosome of the assumptions made in the FE frame model. First, while additional masses, placed onthe first and second floors using steel rails for live loads, slid along the slab slightly during the

Figure 12. Comparison of story displacements for the as-built FE frame model between the experi-mental and simulated responses: (a) t1 time step; (b) t2 time step.

Figure 13. Comparison of peak inter-story drift ratios for the as-built FE frame model between theexperimental and simulated responses under EC 8.

JOURNAL OF EARTHQUAKE ENGINEERING 15

dynamic tests, the masses in the FE frame model were modeled to be fixed (fixed masscondition). Second, the test frame was vibrated under sequential loading scenarios. However,the FE frame model was assumed to be non-damaged under the selected loading scenarios.Finally, to reduce the computational time of the FE frame model, slab, transverse beam, andfoundation elements (noncritical elements), where no significant damage was found from thefull-scale dynamic testing, were simplified with the effective stiffness. These assumptions led tothe slight simulation variation between the experimental and simulated responses. Nevertheless,the FE frame model is able to capture the dynamic responses in terms of the response periods,story displacements, and inter-story drift ratios. In particular, the full-scale dynamic testsdemonstrated that the dynamic responses of the as-built test frame were significantly affected

Figure 14. Comparison of time-history responses for the as-built FE frame model between experimentaland simulated responses under DSP 26: (a) first story; (b) second story.

Figure 15. Comparison of story displacements for the as-built FE frame model between experimentaland simulated responses under DSP 26: (a) t1 time step; (b) t2 time step.

16 J. SHIN ET AL.

by bond-slip behavior in the first-story column bases and exterior beam–column joints underthe DSP 26 loading. Simulations using the bond-slip modeling procedure described hereresulted in maximum displacements in the FE frame model within 10.0% of those observedduring the experimental investigation under DSP 26 loading.

4.3. Retrofitted FE Frame Model

To validate the dynamic responses of the retrofitted FE frame model, the same approachused in the verification works of the as-built FE frame model was adopted. Experimentalresponses of the retrofitted test frame were gathered from the full-scale dynamic test [Shinet al., 2016]. Figure 17 shows the experimental and simulated displacement-time historyresponses of the retrofitted frame under EC 8 loading. As illustrated in Figure 17a and b, theretrofitted FE framemodel captured the experimental results in terms of the response periodin entire time steps. The maximum absolute story displacements at t1 = 1.06 and t2 = 4.40 sare plotted in Figure 18a and b, respectively. The first-story displacement at t1 was under-estimated within approximately 9.8% simulation variation, and the FE frame model under-estimated the first-story displacement at t2 by approximately 6.7%. Figure 19 compares thepeak inter-story drift ratios of the retrofitted frame between experiment and simulation. Themaximum variation for the inter-story drift ratio was estimated within 10.0%.

Figure 20 illustrates the displacement-time history responses under DSP 20, an ultimateloading scenario for the retrofitted test frame. The FE frame model can capture the responseperiods before the beginning of a second loading cycle. However, the response periods wereslightly overestimated during the second loading cycle in DSP 20. This is thought to be due tothe structural damage accumulations induced by the first loading cycle in DSP 20. Figure 21compares the experimental and simulated story displacements at t1 = 6.18 and t2 = 6.23 s. TheFE frame model underestimated the maximum story displacement by approximately 5.5% att1, whereas it overestimated the maximum story displacement by approximately 4.8% at t2.Figure 22 shows the experimental and simulated peak inter-story drift ratios. The peak inter-

Figure 16. Comparison of peak inter-story drift ratio for the as-built FE frame model between experi-mental and simulated responses under DSP 26.

JOURNAL OF EARTHQUAKE ENGINEERING 17

story drift ratio was slightly underestimated by approximately 11.5% compared with theexperimental results in the second story.

4.4. Damage

Figure 23 compares the damage at the first-story column bases between the experimental andsimulated results. Figure 23a shows the locations of damage observed from the full-scaledynamic testing for the as-built and retrofitted test frames. As shown in Figure 23b, diagonalcracks at the lap-splice zone of the first-story column were detected in the as-built test frame.For the retrofitted test frame, a crack at the edge of the column base was observed as shown

Figure 17. Comparison of time-history responses for the retrofitted FE frame model between theexperimental and simulated responses under EC 8: (a) first story; (b) second story.

Figure 18. Comparison of story displacements for the retrofitted FE frame model between experimentaland simulated responses under EC 8: (a) t1 time step; (b) t2 time step.

18 J. SHIN ET AL.

in Figure 23c. Figure 23d and 23e shows the damage on the first-story column lap-splicezones in the as-built and retrofitted FE frame models using effective plastic strains. The colorfringe level for the effective plastic strains of concrete materials is scaled from 0 to 2. Thevalue of 2 in the fringe level represents concrete cracking under tension and concretecrushing under compression [Lin et al., 2014]. As shown in Figure 23d, the as-built FEframe model exhibits diagonal cracks at the first-story column lap-splice zone, similar towhat was observed during the full-scale dynamic experiments. The retrofitted FE framemodel simulates the concrete crack at a gap between the FRP column jacketing system and

Figure 19. Comparison of peak inter-story drift ratio for the retrofitted FE frame model between theexperimental and simulated responses under EC 8.

Figure 20. Comparison of time-history responses for the retrofitted FE frame model between theexperimental and simulated responses under DSP 20: (a) first story; (b) second story.

JOURNAL OF EARTHQUAKE ENGINEERING 19

the foundation (Figure 23e), similar to the observed damage from the full-scale dynamictesting.

4.5. Effects on Bond-Slip Modeling

To investigate the necessity of including bond-slip performance in simulations, thedynamic responses of FE frame models described in the previous sections (referred toas “True bond-slip”) were compared to those of FE frame models with two different otherbond-slip conditions: (1) true bond-slip effects with all good bonding conditions (referredto as “Good bond-slip”), and (2) no bond-slip effects (perfectly bonded between reinfor-cing bars and surrounding concrete, referred to as “No bond-slip”). The true bond-slipmodels have combinations of poor and good bonding conditions based on reinforcingdetailing in the column lap-splice and panel zones as described above. The good bond-slip

Figure 21. Comparison of story displacements for the retrofitted FE frame model between the experi-mental and simulated responses under DSP 20: (a) t1 time step; (b) t2 time step.

Figure 22. Comparison of peak inter-story drift ratio for the retrofitted FE frame model between theexperimental and simulated responses under DSP 20.

20 J. SHIN ET AL.

models were modeled with the good bonding condition in all possible areas where bond-slip effects occur regardless of the reinforcing detailing. Additionally, the no bond-slipmodels deactivated the one-dimensional slide line models in the column lap-splice andpanel zones and the beam elements were merged with the nodes of concrete solid elements(perfect bonding between the reinforcing bars and the concrete material). Those effectswere estimated under each ultimate loading scenario (DSP 26 for the as-built FE framemodel and DSP 20 for the retrofitted FE frame model) because the bond-slip effects aremarginal in elastic range behavior due to negligible slip displacements between thereinforcing bars and the surrounding concrete.

Figure 24 shows the roof time-history responses of the true bond-slip, good bond-slip,and no bond-slip models. Figure 25 compares the maximum simulated responses of the truebond-slip models to those of the no bond-slip models for the first and second sine vibrationsof simulations. For the as-built frame models, as shown in Figure 24a, the overall responsesof the no bond-slip model are significantly less than those of the true bond-slip model. Themaximum response of the no bond-slip model in the first sine vibration is approximately39.0% lower than that of the true bond-slip model (see Figure 25). The main reason for thissignificant difference is that models using no bond-slip effects exhibit perfect bond betweenthe reinforcing bars and the surrounding concrete in the lap-splices and panel zones, whichgreatly exaggerates the overall stiffness of the area.

Figure 23. Comparison of damage between experiment and simulation: (a) damage location of the as-built and retrofitted test frames; (b) damage observed on the as-built test frame [Wright, 2015]; (c)damage observed on the retrofitted test frame [Shin et al., 2016]; (d) damage suggested by the as-builtFE frame model; (e) damage suggested by the retrofitted FE frame model.

JOURNAL OF EARTHQUAKE ENGINEERING 21

The dynamic responses of the good bond-slip model are also less than those of the truebond-slip model. The maximum differences between the true bond-slip and good bond-slip models for the first and second sine vibrations are approximately 22.0% and 17.0%,respectively (see Figure 25). The good bond-slip model was modeled with a good bondingcondition in poor bonding zones, such as first-story lap-splice and panel zones, and thisinappropriate bonding condition resulted in higher bonding stiffness and stresses betweenthe reinforcing bars and the surrounding concrete in the poor bonding zones.

0 2 4 6 8 10 12 14 16 18

-150

-100

-50

0

50

100

150

Time (sec)

Dis

plac

emen

t (m

m)

First sine vibration Second sine vibration

True bond-slipGood bond-slipNo bond-slip

0 2 4 6 8 10 12 14-150

-100

-50

0

50

100

150

Time (sec)

Dis

plac

emen

t (m

m)

First sine vibration Second sine vibration

True bond-slipGood bond-slipNo bond-slip

(a)

Figure 24. Roof time-history responses of the true bond-slip, good bond-slip, and no bond-slip models:(a) as-built FE frame model; and (b) retrofitted FE frame model.

0

10

20

30

40

50

Dif

fere

nce

of m

axim

um r

espo

nses

(%

)

As-built FE frame model Retro fitted FE frame model

First sinevibration

Second sinevibration

First sinevibration

Second sinevibration

True bond-slip vs No bond-slipTrue bond-slip vs Good bond-slip

Figure 25. Comparison of maximum responses for the first and second sine vibrations.

22 J. SHIN ET AL.

Compared with the true bond-slip model for the retrofitted frame, Figure 24b showsthat the overall responses of the no bond-slip and good bond-slip models are slightlyless. In particular, during the first sine vibration of the simulations, the maximumresponses of the no bond-slip and good bond-slip models are approximately 13.0% and8.0% lower than those of the true bond-slip model, respectively. These slight differencesare due to the effectiveness of the FRP column jacket system in the first-story columns,which delayed bond-slip effects by minimizing concrete damage within the first-storylap-splice and panel zones in the early steps of the simulations. In other words, duringthe early run-time, the effects of bond-slip models for the retrofitted FE frame modelsare marginal. However, after the loading step is increased, the dynamic responses of theno bond-slip model were significantly less than those of the true bond-slip model byapproximately 25.0%. This is because the assumption of no-bond slip, in which thereinforcing bars were perfectly bonded with the surrounding concrete in the lap-spliceand panel zones, exaggerated bonding stiffness in all possible bonding zones. Similar tothe no bond-slip model, Figure 25 illustrates that the good bond-slip model has ahigher difference with the true bond-slip model during the second sine vibration of thesimulation than during the first sine vibration because of the bond-slip model withinappropriate bonding conditions in the first-story panel zones. This modeling assump-tion failed to predict reasonable bond-slip behavior during the second sine vibration ofthe simulation.

5. Summary and Conclusion

The present work developed numerical FE models of full-scale as-built (non-ductile) andretrofitted test frames subjected to dynamic loads using LS-DYNA [LSTC – LivermoreSoftware Technology Corporation, 2013]. Past experimental studies [Wright, 2015; Shinet al., 2016] demonstrated that the bond-slip effects significantly affected the soft-storyfailure mechanism of the as-built test frame. Based on these full-scale dynamic tests, thisstudy determined the bonding performance conditions and the failure modes for thelocations where the bond-slip effects were detected. The bond-slip behavior was modeledusing one-dimensional slide line models between the reinforcing bars and the surroundingconcrete. The developed FE frame models were simulated under seismic and sine vibra-tions (ultimate loading scenarios) measured from the full-scale dynamic tests.Additionally, the effects on various bond-slip models were estimated under the ultimateloading scenarios. Based on this investigation, the following conclusions can be drawn.

(1) The developed FE frame models were validated with past experimental responsesunder the seismic and sine vibrations in terms of story displacements and inter-story drifts within 12.0% simulation variation. This slight simulation variation forthe ultimate loading scenarios appropriately verified the experiment-based bond-slip modeling process utilized in this study. The simulation variation was attributedto the following modeling assumptions: (1) fixed mass conditions; (2) non-damagedconditions; (3) effective stiffness using an elastic material model; and (4) assumedfrictional coefficients in the surface layers.

(2) By comparing the roof displacement time-histories of true bond-slip models to thoseof good bond-slip and no bond-slip models, the effects on bond-slip modeling were

JOURNAL OF EARTHQUAKE ENGINEERING 23

estimated. The dynamic responses of no bond-slip and good bond-slip models wereless than those of the true bond-slip models. This is because inappropriate bond-slipmodels produced higher bonding properties between the beam elements and thesurrounding concrete solid elements in the lap-splice and panel zones, where thebond-slip effects occur. These bonding models affected the underestimation of thedynamic responses. Therefore, to capture reasonable simulations for the FE framemodels, appropriate bond-slip modeling in the possible bonding zones is needed.

Acknowledgments

This support is gratefully acknowledged. Any opinions, findings, conclusions, or recommendationsare those of the authors and do not necessarily reflect the views of other organizations.

Funding

This work was supported by the Division of Civil, Mechanical and Manufacturing Innovation[CMMI-1041607].

References

ACI Committee – American Concrete Institute, ACI 318-63. [1963] Building Code Requirements forReinforced Concrete, American Concrete Institute, Detroit, Michigan.

Akguzel, U. [2011] “Seismic performance of FRP retrofitted exterior RC beam-column joints undervarying axial and bidirectional loading,” Ph.D. thesis, Department of Civil Engineering,University of Canterbury, Christchurch, New Zealand.

ASCE – American Society of Civil Engineers, ASCE/SEI 41-13. [2014] Seismic Evaluation andRetrofit of Existing Buildings, American Society of Civil Engineers, Reston, Virginia.

Aschheim, M., Gülkan, P., Sezen, H., Bruneau, M., Elnashai, A. S., Halling, M., Love, J. andRahnama, M. [2000] “Performance of buildings,” Earthquake Spectra 16(S1), 237–279.

Bao, Y., Kunnath, S.K., El-Tawil, S., & Lew, H.S. [2008]. Macromodel-based simulation of pro-gressive collapse: RC frame structures. Journal of Structural Engineering 134(7): 1079–91.

Bao, X. and Li, B. [2010] “Residual strength of blast damaged reinforced concrete columns,”International Journal of Impact Engineering 37(3), 295–308.

Beeby, A. W. [1978] “Corrosion of reinforcing steel in concrete and its relation to cracking,”Structural Engineer 56, 3.

Bracci, J. M., Reinhom, A. M. and Mander, J. B. [1995] “Seismic resistance of reinforced concreteframe structures designed for gravity loads: performance of structural system,” ACI StructuralJournal 92(5), 597–609.

Broadhouse, B. J. [1986] DRASTIC: A Compute Code for Dynamic Analysis of Stress Transients inReinforced Concrete, Safety and Engineering Science Division, AEE, Winfrith.

CEB-FIP MODEL CODE [1990]. “Model code for concrete structures,” In: Comité Euro-International du Béton. Secretariat permanent. Case Postale 88, CH-1015 Lausanne,Switzerland.

Crawford, J. E., Wu, Y., Choi, H., Magallanes, J. M. and Lan, S. [2012] “Use and validation of therelease III K&C concrete material model in LSDYNA,” TR-11-36.6 Technical report, Karagozian& Case, Glendale, California.

Crawford, J. E., Wu, Y., Magallanes, J. M. and Choi, H. J. [2013] “The importance of shear-dilatancybehaviors in RC columns,” International Journal of Protective Structures 4(3), 341–377.

24 J. SHIN ET AL.

Davidson, J. [2008] “Advanced computational dynamics simulation of protective structuresresearch,” Technical Report No. AFRL-RX-TY-TR-2008-4610, Auburn University, Auburn, Ala.

Deaton, J. B. [2013] “Nonlinear finite element analysis of reinforced concrete exterior beam-columnjoints with nonseismic detailing,” Ph.D. thesis, School of Civil and Environmental Engineering,Georgia Institute of Technology, Atlanta, Georgia.

El-Attar, A. G., White, R. N. and Gergely, P. [1997] “Behavior of gravity load designed reinforcedconcrete buildings subjected to earthquake,” ACI Structural Journal 94(2), 133–145.

ElGawady, M., Endeshaw, M., McLean, D. and Sack, R. [2010] “Retrofitting of rectangular columnswith deficient lap splices,” Journal of Composites for Construction 14(1), 22–35.

Engindeniz, M. [2008] “Repair and strengthening of pre-1970 reinforced concrete corner beam-column joints using CFRP composites,” Ph.D. thesis, School of Civil and EnvironmentalEngineering, Georgia Institute of Technology, Atlanta, Georgia.

FEMA – Federal Emergency Management Agency, FEMA 547. [2006] Techniques for the SeismicRehabilitation of Existing Buildings, Federal Emergency Management Agency, Washington D.C.

Haroun, M. A., Mossalam, A. S., Feng, Q. and Elsanadedy, H. M. [2003] “Experimental investiga-tion of seismic repair and retrofit of bridge columns by composite jackets,” Journal of ReinforcedPlastics and Composites 22(14), 1243–1268.

Hu, H. T., Huang, C. S., Wu, M. H. and Wu, Y. M. [2003] “Nonlinear analysis of axially loadedconcrete-filled tube columns with confinement effect,” Journal of Structural Engineering 129(10),1322–1329.

Kwon, M. and Spacone, E. [2002] “Three-dimensional finite element analyses of reinforced concretecolumns,” Computers & Structures 80(2), 199–212.

Lee, D. and Shin, A. H. C. [2016] “Finite element study on the impact responses of concretemasonry unit walls strengthened with fiber-reinforced polymer composite materials,” CompositeStructures 154, 256–268.

Lin, X., Zhang, Y. X. and Hazell, P. J. [2014] “Modelling the response of reinforced concrete panelsunder blast loading,” Materials & Design 56, 620–628.

LSTC – Livermore Software Technology Corporation. [2013] LS-DYNA Keyword User’s ManualVersion 971/R7.0. Livermore Software Technology Corporation, CA.

Luccioni, B. M., López, D. E. and Danesi, R. F. [2005] “Bond-slip in reinforced concrete elements,”Journal of Structural Engineering 131(11), 1690–1698.

Malvar, L. J., Crawford, J. E., Wesevich, J. W. and Simons, D. [1997] “A plasticity concrete materialmodel for DYNA3D,” International Journal of Impact Engineering 19(9), 847–873.

Moradi, L. G. [2007] “Resistance of membrane retrofit concrete masonry walls to lateral pressure,”Ph.D, Dissertation, Univ. of Alabama at Birmingham, Birmingham, Ala.

Mutalib, A. A. and Hao, H. [2010] “Numerical analysis of FRP-composite-strengthened RC panelswith anchorages against blast loads,” Journal of Performance of Constructed Facilities 25(5), 360–372.

Nam, J. W., Kim, H. J., Kim, S. B., Kim, J. H. and Byun, K. J. [2009] “Analytical study of finiteelement models for FRP retrofitted concrete structure under blast loads,” International Journal ofDamage Mechanics 18(5), 461–490.

NEES@UCLA. [2015] Network for Earthquake Engineering Simulation at the University ofCalifornia, Los Angeles, University of California, Los Angeles, Los Angeles, California. http://nees.ucla.edu.

Park, S. and Mosalam, K. M. [2012] “Experimental and analytical studies on old reinforced concretebuildings with seismically vulnerable beam-column joints,” PEER Report 2012/03, PacificEarthquake Engineering Research Center, University of California, Berkeley, Berkeley, California.

Priestley, M. J. N. [1997] “Displacement-based seismic assessment of reinforced concrete buildings,”Journal Earthquake Engineering 1(1), 157–192.

Sause, R., Harries, K. A., Walkup, S. L., Pessiki, S. and Ricles, J. M. [2004] “Flexural behavior ofconcrete columns retrofitted with carbon fiber-reinforced polymer jackets,” ACI StructuralJournal 101(5), 708–716.

JOURNAL OF EARTHQUAKE ENGINEERING 25

Schwer, L. E. and Murray, Y. D. [1994] “A three-invariant smooth cap model with mixed hard-ening,” International Journal for Numerical and Analytical Methods in Geomechanics 18(10),657–688.

Seible, F., Priestley, M. J. N., Hegemier, G. A. and Innamorato, D. [1997] “Seismic retrofit of RCcolumns with continuous carbon fiber jackets,” Journal of Composites for Construction 1(2), 52–62.

Sezen, H., Elwood, K. J., Whittaker, A. S., Mosalam, K. M., Wallace, J. W. and Stanton, J. F. [2000]“Structural engineering reconnaissance of the August 17, 1999 earthquake: kocaeli (Izmit),Turkey Earthquake,” PEER Report 2000/09, Pacific Earthquake Engineering Research Center,University of California, Berkeley, Berkeley, California.

Shi, Y., Hao, H. and Li, Z. X. [2008] “Numerical derivation of pressure–impulse diagrams forprediction of RC column damage to blast loads,” International Journal of Impact Engineering 35(11), 1213–1227.

Shi, Y., Li, Z. X. and Hao, H. [2009] “Bond slip modelling and its effect on numerical analysis ofblast-induced responses of RC columns,” Structural Engineering and Mechanics 32(2), 251–267.

Shin, J., Scott, D. W., Stewart, L. K., Yang, C. S., Wright, T. R. and DesRoches, R. [2016] “Dynamicresponse of a full-scale reinforced concrete building frame retrofitted with FRP column jackets,”Engineering Structures 125, 244–253.

Spacone, E. and Limkatanyu, S. [2000] “Responses of reinforced concrete members including bond-slip effects,” ACI Structural Journal 97(6), 831–839.

Tabiei, A. and Wu, J. [2000] “Roadmap for crashworthiness finite element simulation of roadsidesafety structures,” Finite Elements in Analysis and Design 34(2), 145–157.

Woodson, S. C. and Baylot, J. T. [1999] “Structural collapse: quarter-scale model experiments,”Technical Report SL-99-8, US Army Engineer Research and Development Center, Vicksburg,Mississippi.

Wright, T. R. [2015] “Full-scale seismic testing of a reinforced concrete moment frame using mobileshakers,” Ph.D. thesis, School of Civil and Environmental Engineering, Georgia Institute ofTechnology, Atlanta, Georgia.

Wu, Y.F., & Wei, Y.Y. (2010). Effect of cross-sectional aspect ratio on the strength of CFRP-confined rectangular concrete columns. Engineering Structures 32(1): 32-45.

Wu, Y. and Crawford, J. E. [2015] “Numerical modeling of concrete using a partially associativeplasticity model,” Journal of Engineering Mechanics 141(12), 04015051-1.

Xiao, Y., Wu, H. and Martin, G. R. [1999] “Prefabricated composite jacketing of RC columns forenhanced shear strength,” Journal Structureal Engineering 125(3), 255–264.

Yan, Z. and Pantelides, C. P. [2011] “Concrete column shape modification with FRP shells andexpansive cement concrete,” Constr Build Mater 25(1), 396–405.

Youssf, O., ElGawady, M. A., Mills, J. E. and Ma, X. [2014] “Finite element modelling and dilationof FRP-confined concrete columns,” Engineering Structures 79, 70–85.

26 J. SHIN ET AL.